Build an AI front desk agent for a local business with 5 prompts
A local business misses a call while the owner is under a sink or with a patient. The caller rings the next business on the list. That one missed call can be a lost job worth hundreds.
An AI front desk agent answers every call, chat and text, books the straightforward ones, qualifies the rest and passes anything tricky to a real person. This five-step prompt series designs one for any [niche] in any [city], whether you run the business or set these up for clients. Start at step 1, or read on first.
How to use the series
Run the five prompts in one chat in ChatGPT, Claude or Gemini. Each one fills in your niche, city, services, hours and handoff details, so the output is specific to you. The result is a set of scripts, rules and a system prompt you can paste into whichever voice or chat agent tool you use.
Our worked example: an independent plumbing firm in Manchester with two engineers, emergency cover until 10pm and a shared Google Calendar.
Step 1: The blueprint
Step 1 designs the agent's identity and tone, a table of 8-12 reasons people get in touch, a knowledge sheet built only from your real details, clear handoff rules and guardrails. It ends with a ready-to-paste system prompt. Gaps are marked for the owner to confirm instead of guessed.
In our example: "Water coming through the ceiling" is an instant handoff to the on-call engineer. "How much is a boiler service?" gets the published price and a booking offer.
Step 2: Qualify leads
Step 2 writes a short, natural question flow and scoring rules (hot, warm, not a fit), plus what happens for each score and the summary the owner gets after every conversation. One question per turn, and no blunt budget questions up front.
Step 3: Book without double-booking
Step 3 sets the booking script, how slots are offered, travel buffers, confirmations and reminders, deposit wording and the awkward edge cases like repeat no-shows or out-of-area addresses.
Step 4: Follow up on the ones that got away
Step 4 builds four sequences: missed-call text-back, quote chasers, half-finished bookings and repeat-visit reminders. Each message has timing, channel and a goal, with opt-out lines where the rules need them.
Step 5: Reviews and human handoff
Step 5 asks every customer for a review the same way (no filtering, which breaks most platform rules) and designs a clean handoff to staff, with the note your team receives and a check that someone actually called back. It finishes with a go-live checklist for the whole series.
Want to build the whole thing as an app?
These prompts design the agent's brain. If you'd rather build a full receptionist product, the free AI Receptionist Pro blueprint covers the build. And if you plan to sell front desk agents to local businesses, the AI implementation business series covers choosing a niche and winning clients, and the Offer Stack series helps you package it.
The five steps at a glance
- Blueprint the agent
- Qualify leads
- Book appointments
- Follow up on missed leads
- Request reviews and hand over to staff
Start today
Write down your services, hours and the five most common reasons people call, then run step 1. You'll have a working agent brief before lunch.